Per capita consumption of alcohol has traditionally been considered to be the leading indicator of population levels of alcohol problems. However, some recent research suggests that this relationship may be weakening, and that drinking pattern measures may be preferable to per capita consumption as problem-level indicators. We compared the ability of per capita alcohol consumption and survey-based measures of alcohol use to predict deaths from injuries in Ontario, Canada, for the period 1977-1996. Per capita consumption and percentage of daily drinkers were significantly related to injury mortality, but percentage of drinkers and percentage of episodic heavy drinkers (those who drank five or more drinks on a drinking occasion) were not. Of the measures we examined, per capita consumption was the strongest indicator of mortality rates from injuries. However, the survey-derived measure of percentage of daily drinkers was similar to per capita consumption in ability to predict problem levels.